Hi Janardhan

  1.  Doesn't matter, we can move them around depending on where we want them
  2.  Both should hopefully work fine. I have used some G Colab before really 
liked it, but again i think its up to you.
  3.  sure once we have a notebook we could automate some of the things.
  4.  Yes we have a docker container with systemds but G Colab does not support 
running this [1]. Again the reason i brought it up was because the notebooks 
I've seen was 50%+ setup of systemds.

[1] https://github.com/googlecolab/colabtools/issues/299

Best regards
Sebastian
________________________________
From: Janardhan <[email protected]>
Sent: Wednesday, July 22, 2020 4:38:10 AM
To: [email protected]
Subject: Re: [DISCUSS] open for Jupyter notebook contributions, and end-to-end 
solutions.

Hi all,

This mail would help us structure the notebook files,

Few questions:
1. Which directory should we stage the notebooks, is it
    a. samples/
    b. notebooks/

2. About the notebooks, they contain the notebook examples
to work with [1]
    a. Google Colaboratory,
    b. Databricks platform (mostly MLContext related examples)
in separate folders.

3. Sebastian - can we schedule a github workflow[2]

i.e., step 1: notebook.ipynb (committed to systemds)
       step 2: notebook.html (generated by github workflow)
       step 3: notebook.html (committed to a `deploy` branch in
systemds-website)[3]
       step 4: we will commit this file to our svn repo for
systemds.apache.org

4. Sebastian - you have mentioned something about docker in
this context,
    a. Is it about providing a jupyter server via a port?

[1] https://github.com/j143/notebooks
[2] https://github.com/j143/notebooks/pull/10
[3] https://github.com/j143/systemds-website/pull/8

Thank you,
Janardhan

On Thu, Jul 2, 2020 at 10:10 AM Janardhan <[email protected]> wrote:

> Hi all,
>
> Now the SystemDS project scope has broadened from just ML to complete Data
> Science life cycle,
> in order to showcase our functionality, we are finding use cases for each
> of the steps such as cleaning [1],
> processing [2], and deploying on the cloud with end-to-end solutions (on
> AWS or Google Cloud).
>
> [Note: OneDesign sponsoring $300 worth cloud resource if you are
> developing a solution based on
> AWS cloud with SystemDS also support in building cloud architecture valid
> till July 2020,
> reach out to the author of this mail :) ]
>
> We would like to keep the project open for notebook contributions, the
> notebooks for both ML
> and data processing use cases. (Idea borrowed from Apache Beam &
> TensorFlow tutorials).
>
> Bonus: These notebooks have Google Colab compatibility to work on without
> any configuration.
>
> *Questions:*
> 1. Who will take ownership?
> All the contributors are supporting this feature, in their respective
> components.
>
> 2. Do you have a concrete implementation?
> Yes, In fact, we have tested it.
> Sample notebooks  (Useful for researchers/engineers, start prototyping in
> just 3 minutes)
> a. Algorithms dev:
> https://colab.research.google.com/github/j143/notebooks/blob/master/systemds_dev.ipynb
> b. Deep learning:
> https://gist.github.com/j143/df1fdea505df2c662b326bd689bf5a0d
> c. SystemML library:
> https://github.com/apache/systemds/tree/branch-1.2.0/samples/jupyter-notebooks
>
> 3. Will you mentor and review the PR?
> Yes, we will mentor[3] the contributors and review the PRs in notebooks,
> on request.
>
> [1] https://github.com/apache/systemds/pull/981
> [2]
> https://github.com/apache/systemds/commit/8cbc85a949b3699cde8ed3cf3e3abec6a27fbc60
> [3] https://community.apache.org/mentoringprogramme.html
>
> Thank you,
> Janardhan
>

Reply via email to